1. Field of Invention
The present general inventive concept pertains to semiconductor production, and more particularly, to a method of producing a low cost, high-efficiency, high yield semiconductor useful in numerous applications, for example in a photovoltaic cell or a light-emitting diode.
2. Description of the Related Art
The use of silicon and aluminum as precursors for making a variety of semiconductor members for use in products such as solar cells, light-emitting diodes, high energy energetic materials, and other semiconductor and ceramic applications, requires that such precursors be free of contamination with materials and elements with which these materials are naturally very reactive, such as oxygen, carbon, nitrogen, and iron. Such contamination often creates undesirable effects in the resultant semiconductor member, such as low thermal conductivity, low electrical output, low strength, poor kinetics, and poor energy output or reactivity. Because the bonds created by such contaminations are typically covalent bonds, such contaminants are difficult to remove once they have joined with the precursor. In many cases, the contaminant itself is difficult to measure and the negative effects created by the contaminant, though understood and recognized, are difficult to quantify.
As an example, in the manufacture of photovoltaic cells, commonly known as “solar cells,” generally the solar cells are fabricated using a semiconductor wafer substrate using semiconductor processing techniques to form a p-n junction near a surface of the substrate. Solar radiation impinging on the surface of the substrate creates electron and hole pairs in the bulk of the substrate, which migrate to p-doped and n-doped regions in the substrate, thereby generating a voltage differential between the doped regions. The doped regions are coupled to a conductor on the solar cell to direct an electrical current from the cell to an external circuit coupled thereto. Commonly, the semiconductor wafer substrate is a silicon wafer, with the n-type junction on the surface of the silicon wafer based on phosphorous doping and the p-type junction in the body of the silicon wafer based on boron doping.
In terms of desirability for producing a low cost, high-efficiency, high yield semiconductor useful in manufacturing solar cells, it is recognized that use of solar cells in electrical energy production is attractive, in part due to the inexpensive nature of solar radiation used to fuel energy production in the solar cells. As the fuel costs associated with more conventional means of electrical energy production increase, demand for solar cells for electrical energy production also increases. However, previous methods for production of semiconductors useful in manufacturing solar cells have proven to be labor intensive, energy intensive, and materials intensive, and such previous methods have often resulted in the production of solar cells having relatively low efficiency in electrical energy production. For example, in one known method of producing a solar grade semiconductor wafer, a polycrystalline silicon material is used as the substrate. In this method, the polycrystalline silicon wafer is obtained by first placing and packing essentially pure lumps of silicon in a crucible. The crucible is loaded into a vacuum furnace that has heating elements made of graphite. The vacuum furnace heats the silicon lumps, causing them to melt. Thereafter, the melted silicon is cooled to encourage the formation of a large silicon ingot defining polycrystalline crystal formations. The crucible is commonly constructed of fused silica and, because of the processing temperature of the vacuum furnace, the fused silica partially converts to crystobalite during the melting process within the vacuum furnace. Thus, the crucible is a single use and expensive tool. The melting process also produces a polycrystalline silicon ingot having a large outer volume contaminated by impurities reacted with the melted silicon. This outer volume is cut away and discarded following manufacture of the polycrystalline ingot. Thereafter, the somewhat more pure remaining inner portion of the ingot is thin-cut into discreet polycrystalline silicon wafers using a wire saw in a diamond slurry, causing a very large additional loss of silicon, and also limiting the minimum thickness the wafer can be fabricated. The resultant polycrystalline silicon wafers can then be laminated to a conductive layer, thereby forming the solar cell as described above.
The length of the heating-cooling cycle in the above-described method is often 45-60 hours. Thus, the time associated with performing the above-described heating-cooling cycle can result in significant delays in the production process. One factor in determining the length of the heating-cooling cycle is the time required to heat the silicon feedstock sufficiently to cause it to melt. As a general rule, the smaller the initial lumps of silicon to be melted, the faster the heating and melting cycle. However, it has been found that the process of diminution of the silicon lumps into silicon granules often smears undesirable contaminants onto the surface of the resulting silicon granules. Thus, the silicon lumps used are typically very coarse, having an average size of approximately thirty (30) millimeters. Use of such coarse materials helps to preserve purity in the melted silicon. However, the packing density of the above-discussed silicon lumps in the mold is approximately thirty-five (35) percent of perfect packing, which is significantly less than ideal. Consequently, heat is not conducted efficiently through such a feedstock, and additional heating time is required. Given that (1) the heating elements are on the outside of the crucible, (2) the silicon lumps have relatively little physical contact with one another, and (3) the silicon lumps often serve to “shadow” one another from thermal radiation very heavily; most of the heating occurs by thermal radiation that is accomplished in succession, wherein a relatively exposed silicon lump is heated, and that lump then radiates heat to one or more successive relatively unexposed lumps, i.e., one or more of the silicon lumps that are “shadowed.” In many instances, a partial pressure of argon gas is used to assist in transferring heat to the silicon feedstock.
The above-discussed process of melting the silicon material often results in undesirable random crystal structures present in the silicon ingot, contributing to poor performance of the semiconductor polycrystalline silicon wafer.
For example, in use of the polycrystalline silicon wafer for the manufacture of solar cells, such random crystal structures often result in a low efficiency in the resulting polycrystalline silicon wafer for converting sunlight into electrical energy. Furthermore, during the heating and subsequent cooling of the silicon, two impurities, i.e., silicon carbide (SiC) and dissolved oxygen complexes (including silicon-oxygen complexes), are produced in the silicon feedstock. These impurities cause a reduction in the yield of usable silicon crystal wafers that can be as high as approximately forty (40) percent. Also, these impurities cause additional defects in the crystal structure that further reduce the efficiency and life of the semiconductor polycrystalline silicon wafer.
At least a few factors encourage the synthesis of these impurities. First, the high temperatures achieved in the furnace promote the oxidation of its graphite heating elements by reduction of the fused silica with which the graphite is in physical contact, thus creating a partial pressure of CO and CO2. Other components of the vacuum furnace may be composed of graphite as well, including the insulation material, and likewise, may too be oxidized. This oxidation-reduction reaction commonly yields two gases: carbon monoxide (CO) and carbon dioxide (CO2). These gases then react with the silicon feedstock in the mold to yield silicon carbide and dissolved-oxygen complexes. Second, although rebonded fused silica is a highly refractory substance, it is permeable by carbon oxide gases (e.g., CO and CO2). Thus, carbon oxide gases access the silicon feedstock by permeating the mold. Third, the packing density of the silicon lumps results in spaces that can be permeated and/or occupied by the carbon oxide gases. The surfaces of the silicon lumps that border these spaces serve as additional loci for the oxidation-reduction reaction that yields silicon carbide and dissolved-oxygen complexes. It has further been found that, in the above-referenced polycrystalline silicon production process, the melted silicon within the crucible contains approximately 350 ppm iron oxide (Fe2O3) in solution. At temperatures between approximately 1,100 and 1,600 degrees Centigrade of the melted silicon, crystobalite crystals precipitate within the melted silicon, thereby forcing the iron oxide into the grain boundaries of the silicon as the viscosity of the silicon lowers. Such iron oxide contaminants significantly reduce the carrier life time and efficiency of the resultant polycrystalline silicon wafer.
If a silicon wafer body is uniformly doped at low levels with phosphorous, use of boron to dope the surface of the silicon wafer to establish a p-type junction results in greatly increased efficiency in converting photons to electrons by the resultant solar cell. However, prior art doping technology makes this type of uniform doping of phosphorous in a solar grade silicon wafer impractical in a commercial setting. Specifically, because of the process time of the vacuum furnace and the size of the melted silicon ingot, it is impractical to directly “dope” the melted silicon ingot to form the body of the n-type or p-type junction. Because of the limits of the doping technology, doping of the silicon is generally limited to using boron in the body of the silicon wafer to make the p-type junction and phosphorous at the surface of the silicon wafer to make the n-type junction.
However, the methods of applying phosphorous to the surface of the silicon result in much larger coatings than are needed or can be achieved with boron. Finally, it must be acknowledged that vacuum furnaces generally do not create perfect vacuums, allowing atmospheric gases and potentially other gases to enter. Atmospheric gases include oxidizing agents that, as described previously, can result in the production of impurities.
A further yield loss is incurred by the sawing and slicing of the billet into wafers. Polycrystalline silicon is a relatively hard and brittle material, and thus, the operation of cutting the polycrystalline material is inherently difficult and labor intensive and results in a high mortality rate of the thin-cut silicon wafers due to fracture of the wafers during tooling and handling. In at least some instances, by the time the above-discussed contaminated outer layer is removed from the silicon ingot and the ingot is sliced down to silicon wafers having a thickness of approximately 150-200 microns, and by the time resultant fractured wafers are discarded, the yield of usable thin-cut silicon wafers on starting silicon can be as low as 10-30%.
In use of silicon semiconductor wafers in the manufacture of solar cells, while a silicon wafer of approximately 180 microns in thickness captures substantially all sunlight, a silicon wafer of approximately 40 microns still captures most sunlight, i.e., in excess of 96% of sunlight in normal conditions. The slight loss of light capture of the 40 micron wafer is more than compensated for in efficiency increases by decreased instances of recombination associated with a shorter path through the 40 micron silicon wafer. Thus, it is believed that a thinner wafer, perhaps 40 microns thick, would be optimal for use in solar cells. However, such a wafer cannot be made and handled by prior existing technology absent significant breakage of the wafer as discussed above. In light of the above, the low yield of usable silicon wafer material and the high costs per unit of solar conversion efficiency associated with manufacture of solar cells using the above-discussed process have made use of solar cells manufactured by the above-discussed process for electrical energy production in the residential, commercial, and utility sectors impractical in many applications from an economical point of view without large subsidies from governments and the like.
U.S. Pat. No. 7,604,696 (“the '696 patent), which was issued to Carberry, describes one method of converting a silicon powder slurry into thinly molded solar grade silicon wafer. However, this and other known methods often lead to increased contamination of the resultant silicon wafers during the heating and cooling cycle. Accordingly, there is a continuing need for an improved method to produce a high-efficiency, solar cell.
Other uses for semiconductor members include the manufacture of light-emitting diodes (“LED's”). It is known that the lifetime of a semiconductor used in an LED is largely defined by its operating temperature. Specifically, the lower the operating temperature, the lower the rate of diffusion among the various thin layers of materials which enable the function of the semiconductor, such diffusion being the eventual failure mode of most semiconductors. Thus, high-purity aluminum nitride, which has a high thermal conductivity and a coefficient of thermal expansion similar to silicon, has been found to be beneficial in for use as a semiconductor in the manufacture of LED's. However, aluminum nitride of sufficient purity to achieve the requisite high thermal conductivity has traditionally suffered from a prohibitively high manufacture cost.
Two prior art processes for manufacturing aluminum nitride, carbo-thermal reduction and direct nitridation, each suffer from poor precursors and multiple processes that make it impossible to achieve low cost and low oxygen content of the aluminum nitride product. Aluminum precursors are commonly available in two forms, nanophase precursors and milled aluminum. Particles comprising a nanophase precursor have too high a surface area and too small a size to be easily useful for fabricating ceramic preforms directly. Milled aluminum is commonly available in five micron thick ribbons which are also difficult to work with in classic forming technologies for making ceramic performs. In both nanophase precursors and milled aluminum, and with both the nitriding and carbo-thermic reduction processes, the oxygen content of the finished aluminum nitride is too high to achieve sufficient purity of the aluminum nitride product for beneficial use in an LED. Likewise, silicon nitride suffers in its performance as a semiconductor largely from iron in the grain boundaries of the silicon nitride, which can only be eliminated by having a silicon powder sufficiently free of oxygen and iron as discussed above.
In light of the above, it is desirable to develop a method for manufacturing semiconductor members in which precursors are produced having a high surface area, nearly equaxed morphology, and low presence of contaminants such as oxygen, iron, nitrogen, carbon, and the like, and that such precursors have a particle size and particle size distribution suitable for forming and processing from an un-oxidized, uncontaminated state. It is further desirable that the method allow for manufacture of the semiconductor member while maintaining a heightened purity of the precursors, thus producing a high-purity finished semiconductor product.